Dr. Don Hendricks
-3-
August 28, 1972
and express this as a proportion:
Standard error as proportion = ——- = ——= =
so that approximate confidence limits can be expressed as (for accuracy
one might use t-variate as multiplier for the smaller sample sizes, but the
present procedure seems adequate for making judgements; also we will later
bring)the effect of skewness on confidence limits):
Confidence limits as proportion of mean =
+2 SE(prop.) = + 2
An
Thus if we know or assume the value of the coefficient of variation appropriate for the Atoll, one can calculate the sample sizes necessary for
statements such as the following:
"The probability that the true mean concentration of plutonium in soil
samples from Japtan Island wil] be within + 5% of the observed mean determined
from n samples is .95".
In other words, we can specify a percentage
interval within which we can have high confidence that the actual concentration will fall, given an advance estimate of C (the coefficient of variation).
Note that all we need is C, since the actual concentrations do not enter the
calculations -- we make the statements as proportions or percentages.
Some actual estimates of C for soil plutonium are as follows:
Samples
Locations
Size
Bikini (page 12 of NVO-97/ Summary report
Coefficient
of variation
1
74
12
-42
of the 1969 and 1970 Bikini Surveys)
Eniwetok (Data on samples from Janet, by phone
from 0. Lynch)
Eniwetok (9 composites of 3 each from Phase IIA
9 -
-75 (1.30)*
islands; Alice to Edna)
Rocky Flats (HASL-235)
Nevada Test Site (GMX study)
*adjusting for composites
33°
~ Stratum I
3.6
10
.66
Stratum I]
4
55
- Stratum IV
6
1.36